
Data and codes for the empirical results in 
"A Unifying Approach to the Empirical Evaluation of Asset Pricing Models"
by Francisco Pe�aranda and Enrique Sentana.

In particular, we report empirical evaluations of the following models:

- Table 1: CAPM.

- Table 2: CCAPM.

- Table A1: Epstein-Zin model.

- Table A2: CCAPM with a gross return.

We use the data in Lustig and Verdelhan (2007), where the payoffs to price are the annual excess returns on 8 currency portfolios (1953-2002).
Our codes are written for Gauss.


--------------------------------------


DATA

Files: CPI.txt, FF3a.txt, LV_fac.txt, and LV_port.txt

The files LV_fac and LV_port correspond to the data in Lustig and Verdelhan (2007), available in Adrien Verdelhan's web page:

- LV_fac has 2 pricing factors, Rvwexc and Consumption Growth.

- LV_port has 8 portfolios to be priced.

The files CPI and FF3a are used to construct the gross return in Table A2, which is not priced in Lustig and Verdelhan (2007):

- FF3a contains the 3 Fama-French factors and the riskfree rate, from Kenneth French's web page. We only use the last column.

- CPI contais the consumer price index, from the US Bureau of Labor Statistics. It is used to translate the riskfree rate into real terms.


--------------------------------------


CODES

Files: table1.prg, table2.prg, tableA1.prg, and tableA2.prg

Each code corresponds to a table in the paper as the file's name indicates.
Each code is self-contained. They only need to read the data and the Gauss library "optmum".

We implement each SDF or regression method by continuously updated (CU), iterated and two-step (2S) GMM.
Each one of the three variants is associated to one of the three columns in each table.
The parameter "n_iter" at the beginning of the code controls which variant to run: 1 for CU-GMM, 2 for 2S-GMM, more for iterated GMM.
Once a value of "n_iter" is chosen, the code can be run and all the numbers of the corresponding column will be obtained.

The convergence of the estimates and the J statistic should be checked when running iterated GMM. 
To check the evolution of the estimates and the J statistic, simply delete @ in these parts of the code:  

@
    "--------------";
    "iteration" i ;
    "j, retcode and theta1'"; x[1,.];
@

In the output of the two symmetric normalizations, we report an equivalent angle below the estimate in case its value is outside (-pi/2,+pi/2).


--------------------------------------


EXAMPLE

This file is the output from running file "table1" with n_iter=1, which yields the first column of Table 1 in the paper.



